Meta Cracks Down on Excessive AI Tool Usage Amid Staggering Compute Token Costs

Meta is facing an unexpected expense in its use of internal AI tools. According to a report, employees have been consuming an excessive amount of 'compute tokens,' which are essentially units of text processed by AI models. These tokens have nothing to do with blockchain or cryptocurrency.

The numbers are staggering: Meta employees collectively consumed 60 trillion tokens in a single 30-day period. Some individual users burned through hundreds of billions of tokens on their own, far exceeding the average usage of a few thousand tokens per conversation.

To manage this cost, Meta is implementing strict limits on token usage. The company will essentially cap how much AI each employee can consume, effectively putting an end to what was described as a 'competitive addiction' to large language model usage.